AN EVALUATION OF CHIDAMBER AND KEMERER'S LACK OF COHESION IN METHOD (LCOM) METRIC USING DIFFERENT NORMALIZATION APPROACHES

Chidamber and Kemerer first defined a cohesion measure for object-oriented software – the Lack of Cohesion in methods (LCOM) metric. However, the LCOM metric is not a normalized measure, hence the measure does not satisfy all the properties for cohesion measures. This paper presents three approaches which could be used to normalize the LCOM metric. The approaches are the Sigmoid LCOM (SLCOM) normalization, the Bowles LCOM (BLCOM) normalization and the Best-Fit LCOM (BFLCOM) normalization. These approaches were evaluated using field data from a Java based industrial system. Class cohesion was measured using the LCOM metric. The three normalization approaches were applied to the measured LCOM metric values. The result of the study suggests that the Sigmoid LCOM may not be a good transformation for the LCOM metric, the Bowles and Best-fit normalized LCOM measures may be used with the LCOM, while the Best-fit LCOM seem to be the best normalization approach for the LCOM metric.